AI Workshop at DEIF incl. Factory Tour

After a welcome by Holger Thater, vice-president of tekom Danmark, Majken Brunsborg Lauritsen and Jeanine Shepstone gave an insights into the technical writing setup at DEIF. We also had the chance to see the state-of-the-art DEIF production facilities, which was very interesting and impressive. The day ended with a workshop with 6 micro-presentations focused on using AI in technical documentation.

A big thank you for DEIF for hosting the event - and a shout-out to the 6 "micro-presenters" for being bold enough to step out on the podium!

Attached, you can find the presentations given at the event (only accessible for tekom members). To eat our own medicine, we added the raw AI workshop transcript that was automatically recorded and the resulting AI notes - see for yourself and draw your own conclusion on if that is a use case you want to explore yourself deeper.

Here’s a summarized version of the meeting notes in 15 sentences:

1. The team uses AI to summarize topics by copying them to an internal chatbot, which generates a one-sentence summary to be used in a <shortdesc) element in a DITA topic that is well-received.
2. Despite integration challenges, the manual copying of summaries from CCMS to AI and back to the CCMS still remains beneficial and easy to implement.
3. AI translation struggles significantly with Eastern European languages, resulting in poor translation quality and formatting issues, especially with PDFs.
4. There is an internal push to use AI for translations to save costs, but this leads to compliance issues since translations are not always up to standard.
5. Copilot helps manage emails effectively by recalling past conversations and providing relevant information, making email management more efficient.
6. AI is used to create engaging meeting agendas and structure meetings effectively, ensuring both engagement and desired outcomes.
7. AI creates job descriptions by providing specific fields and relevant covers, proving to be effective.
8. AI faces limitations in Excel data manipulation, particularly with unformatted data, making manual formulas and macros more efficient.
9. AI was used to answer HR interview questions accurately based on past conversations.
10. Different AI models are preferred for specific tasks, such as Gemini for writing definitions and ChatGPT for creating images.
11. Company policies limit experimentation with various AI models for different tasks.
12. AI models that produce natural-sounding text are preferred for tasks requiring human-like language generation.
13. AI struggles with large Excel sheets and unformatted data but is used for tasks like replacing material numbers.
14. A cautionary tale was shared about AI recommending competitor products, highlighting the need for proper control and oversight.
15. Follow-up tasks include investigating AI translation issues, exploring integration options for summarization, evaluating AI use in Excel, utilizing AI for job descriptions, assessing AI model suitability, and using AI for meeting planning.

On behalf of the tekom Danmark board

Holger Thater